netfilter: ipv6: move POSTROUTING invocation before fragmentation
authorJan Engelhardt <jengelh@medozas.de>
Tue, 13 Apr 2010 13:28:11 +0000 (15:28 +0200)
committerPatrick McHardy <kaber@trash.net>
Tue, 13 Apr 2010 13:28:11 +0000 (15:28 +0200)
Patrick McHardy notes: "We used to invoke IPv4 POST_ROUTING after
fragmentation as well just to defragment the packets in conntrack
immediately afterwards, but that got changed during the
netfilter-ipsec integration. Ideally IPv6 would behave like IPv4."

This patch makes it so. Sending an oversized frame (e.g. `ping6
-s64000 -c1 ::1`) will now show up in POSTROUTING as a single skb
rather than multiple ones.

Signed-off-by: Jan Engelhardt <jengelh@medozas.de>
Signed-off-by: Patrick McHardy <kaber@trash.net>
net/ipv6/ip6_output.c

index 4535b7a..236ac78 100644 (file)
@@ -82,22 +82,6 @@ int ip6_local_out(struct sk_buff *skb)
 }
 EXPORT_SYMBOL_GPL(ip6_local_out);
 
-static int ip6_output_finish(struct sk_buff *skb)
-{
-       struct dst_entry *dst = skb_dst(skb);
-
-       if (dst->hh)
-               return neigh_hh_output(dst->hh, skb);
-       else if (dst->neighbour)
-               return dst->neighbour->output(skb);
-
-       IP6_INC_STATS_BH(dev_net(dst->dev),
-                        ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
-       kfree_skb(skb);
-       return -EINVAL;
-
-}
-
 /* dev_loopback_xmit for use with netfilter. */
 static int ip6_dev_loopback_xmit(struct sk_buff *newskb)
 {
@@ -111,8 +95,7 @@ static int ip6_dev_loopback_xmit(struct sk_buff *newskb)
        return 0;
 }
 
-
-static int ip6_output2(struct sk_buff *skb)
+static int ip6_finish_output2(struct sk_buff *skb)
 {
        struct dst_entry *dst = skb_dst(skb);
        struct net_device *dev = dst->dev;
@@ -150,8 +133,15 @@ static int ip6_output2(struct sk_buff *skb)
                                skb->len);
        }
 
-       return NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING, skb, NULL, skb->dev,
-                      ip6_output_finish);
+       if (dst->hh)
+               return neigh_hh_output(dst->hh, skb);
+       else if (dst->neighbour)
+               return dst->neighbour->output(skb);
+
+       IP6_INC_STATS_BH(dev_net(dst->dev),
+                        ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
+       kfree_skb(skb);
+       return -EINVAL;
 }
 
 static inline int ip6_skb_dst_mtu(struct sk_buff *skb)
@@ -162,21 +152,28 @@ static inline int ip6_skb_dst_mtu(struct sk_buff *skb)
               skb_dst(skb)->dev->mtu : dst_mtu(skb_dst(skb));
 }
 
+static int ip6_finish_output(struct sk_buff *skb)
+{
+       if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
+           dst_allfrag(skb_dst(skb)))
+               return ip6_fragment(skb, ip6_finish_output2);
+       else
+               return ip6_finish_output2(skb);
+}
+
 int ip6_output(struct sk_buff *skb)
 {
+       struct net_device *dev = skb_dst(skb)->dev;
        struct inet6_dev *idev = ip6_dst_idev(skb_dst(skb));
        if (unlikely(idev->cnf.disable_ipv6)) {
-               IP6_INC_STATS(dev_net(skb_dst(skb)->dev), idev,
+               IP6_INC_STATS(dev_net(dev), idev,
                              IPSTATS_MIB_OUTDISCARDS);
                kfree_skb(skb);
                return 0;
        }
 
-       if ((skb->len > ip6_skb_dst_mtu(skb) && !skb_is_gso(skb)) ||
-                               dst_allfrag(skb_dst(skb)))
-               return ip6_fragment(skb, ip6_output2);
-       else
-               return ip6_output2(skb);
+       return NF_HOOK(NFPROTO_IPV6, NF_INET_POST_ROUTING, skb, NULL, dev,
+                      ip6_finish_output);
 }
 
 /*